Abstract: A computer-implemented method, system and computer program product for generating visualizations for semi-structured data. Visualization data is extracted from infographics depicting semi-structured data. The visualization data that is extracted includes the traits or characteristics of the semi-structured data depicted in the infographics (e.g., dimension), the characteristics of the infographics (e.g., location of the depicted data), and the constraints or display requirements (e.g., display target value in a particular axis). A trait and constraint rule set is then generated based on the extracted visualization data. The trait and constraint rule set includes a set of rules that maps the display requirements to the particular set of traits or characteristics exhibited by the semi-structured data displayed in the infographics.

Abstract: Methods, computer program products, and/or systems are provided that perform the following operations: obtaining a series of indicator diagrams corresponding to strokes of a pumpjack over a specific time duration, dividing each indicator diagram into a plurality of location segments in a direction of location of the rod; obtaining load difference features between upstroke loads and corresponding downstroke loads in the plurality of location segments; identifying a location segment with an abnormal load difference feature based on a time series data of load difference feature corresponding to one of the plurality of location segments, the time series data of load difference feature including a series of data points of load difference feature of the one of the plurality of location segments in time order; and providing an indication of a potential problem based, at least in part, on the identification of the location segment with an abnormal load difference feature.

Abstract: Using incremental learning techniques and statistic-based approaches to update models. In some instances, a set of residual values are calculated by extracting features by determining the difference between an original data model and a new data model. The calculated set of residual values are ultimately used to update the original model to produce a final data model that is structured and configured to seamlessly process data from both the original data model and the new data model.

Abstract: According to embodiments, a method, a device and a computer program product for model update determination are proposed. In the method, a plurality of historical data items and a plurality of new data items are obtained. The plurality of historical data items were used for training a model, and the plurality of new data items are to be applied to the model. At least one of an overall difference, a structural difference, and a confidence difference between the plurality of historical data items and the plurality of new data items is determined. Thereby, an indication indicating whether to update the model is determined based on the at least one of the overall difference, the structural difference, and the confidence difference.

Abstract: Embodiments of the present disclosure relate to methods, systems, and computer program products for classification based on imbalanced dataset. In a method, a plurality of predictive models are generated based on a plurality of positive samples included in an imbalanced dataset, respectively, an amount of the plurality of positive samples being below an amount of a plurality of negative samples that are included in the imbalance dataset, and each sample in the plurality of positive and negative samples including a plurality of parameters. A plurality of influential parameter groups are identified from the plurality of parameters for the plurality of positive samples, respectively. A final predictive model is determined based on the plurality of predictive models and the plurality of influential parameter groups, the final predictive model being used for classifying a sample into one of a positive type and a negative type.

Abstract: Embodiments herein describe computer-implemented methods, computer program products, and computer systems for determining record importance. The methods may include providing a first trained model having a first trained model accuracy. Further, the methods may include clustering the training data to generate clustered data groups, extracting a first clustered data group from the clustered data groups to identify first model test data, processing the first model test data using the first trained model to generate first trained model output data having first test data accuracy, and labeling the first clustered data group with a first record importance level based on a first comparison between the first trained model accuracy and the first test data accuracy. Further, the methods may include clustering the training data by processing the training data using hierarchical clustering to group the training data into the clustered data groups based on features corresponding to a hierarchy of importance.

Abstract: Performing predictive analysis on running batch jobs is provided. A series of batch end time predictive models is retrieved according to a sequence of milestone jobs in a batch of jobs. Retrieved batch end time predictive models are assembled into an aggregate batch end time predictive model to increase accuracy and stability of an end time prediction of the batch of jobs. The aggregate batch end time predictive model is utilized to predict an end time of the batch of jobs during running of the batch of jobs to form a predicted end time of the batch of jobs.

Abstract: An approach is provided in which the approach constructs a 3-dimensional (3D) matrix based on a plurality of historical transactions performed by a user. The 3D matrix includes a set of features, a set of rows, and a set of channels. The approach trains a convolutional neural network using the 3D matrix, and then uses the trained convolutional neural network to predict a risk level of a new transaction initiated by the user. The approach transmits an alert message based on the predicted risk level.

Abstract: An approach is provided in which the approach constructs a 3-dimensional (3D) matrix based on a plurality of historical transactions performed by a user. The 3D matrix includes a set of features, a set of rows, and a set of channels. The approach trains a convolutional neural network using the 3D matrix, and then uses the trained convolutional neural network to predict a risk level of a new transaction initiated by the user. The approach transmits an alert message based on the predicted risk level.

Abstract: A computer-implemented method, system and computer program product for detecting feature traits from an imbalanced dataset. A first regression model is built based on a simulated dataset to compute contribution scores for the features to make a target a positive case. A variance in the features' original values for each feature of the first set of features (those features with contribution scores for positive cases that exceed a threshold value) between the positive and negative cases is determined. A second regression model is built to calculate a predictor importance value for a second set of features (features from the first set of features with a variance in their original values for both positive and negative cases that exceeds a threshold value). Feature traits are then extracted from a group of clustered positive cases with features of the second set of features containing a predictor importance value exceeding a threshold value.

Abstract: In an approach for detecting and profiling influential predictors in a machine learning model, a processor receives a case as an input to a pre-trained model for a prediction, the prediction being based on a plurality of predictors. A processor determines a range for a predictor where the predictor varies without changing the prediction. A processor generates a data set by sampling data across the determined range of the predictor. A processor builds a predictive model with the generated data set. A processor outputs a prediction result including an insight about the range of the predictor.

Abstract: A programmable device includes a functional module, a pre-allocation manager, a first connection line, and a second connection line, wherein the pre-allocation managers are connected by the first connection lines, and the pre-allocation managers are connected to the functional modules by the second connection lines; the first connection lines are used for data transmission between the pre-allocation mangers, and a transmission direction is determined according to the configuration; the second connection lines are used for data transmission between the pre-allocation managers and the functional modules; the pre-allocation mangers are used for data transmission between the first connection lines and for data transmission between the first connection lines and the functional modules. The first connection lines are configured as connection line segments for transmission in both directions, and a wiring structure is designed in a direction and shape meeting wiring requirements.

Abstract: Embodiments relate to a system, computer program product, and method for determining missing values in respective data records with an explanatory analysis to provide a context of the determined values. Such method includes receiving a dataset including incomplete data records that are missing predictors and complete data records. A model is trained with the complete data records and candidate predictors for the missing predictors are generated. A predictor importance value is generated for each candidate predictor and the candidate predictors that have a predictor importance value in excess of a first threshold value are promoted. Respective promoted candidate predictors are inserted into the respective incomplete data records, thereby creating tentative data records. The tentative data records are injected into the model, a fit value is determined for each of the tentative data records, and a tentative data record with a fit value exceeding a second threshold value is selected.

Abstract: An approach is provided in which the approach builds a combination model that includes a normal status model and an abnormal status model. The normal status model is built from a set of time-sequenced normal status records and the abnormal status model is built from a set of time-sequenced abnormal status records. The approach computes a set of time-sequenced coefficient combination values of the normal status model and the abnormal status model based on applying a set of fitting coefficient characteristics to the normal status model and the abnormal status model. The approach performs goal seek analysis on a system using the combination model and the set of time-sequenced coefficient combination values.

Abstract: A multiview backlight, multiview display, and method of multiview backlight operation include reflective multibeam elements having one or more curved reflective surfaces configured to provide emitted light having directional light beams with directions corresponding to view directions of a multiview image. The multiview backlight includes a light guide configured to guide light and an array of the reflective multibeam elements. Each reflective multibeam element includes a plurality of reflective sub-elements and is configured to reflectively scatter out a portion of the guided light as the emitted light. The multiview display includes the multiview backlight and an array of light valves to modulate the directional light beams to provide the multiview image. A reflective sub-element of the reflective sub-element plurality includes a curved reflective surface, a surface curvature of the curved reflective surface being in a plane parallel to a guiding surface of the light guide.

Abstract: A method of clustering data generated by an unknown model, the method including accessing the data, wherein the data includes a prediction target and a confidence, extracting a data group with high prediction confidence from the data, wherein the data group comprises a plurality of data cases, and where each of the data cases is described by a plurality of predictors, identifying high rank predictors of each the data cases in the data group, transforming the data group into a transformed data group including only the high rank predictors for each of the data cases, wherein the high rank predictors are ranked within each of the data cases included in the transformed data group, clustering the transformed data group to generate clusters, and profiling the clusters to extract an insight about the unknown model.

Abstract: A micro-slit scattering element based backlight, a multiview display, and a method of backlight operation include reflective micro-slit scattering elements configured to provide emitted light having a predetermined light exclusion zone. The micro-slit scattering element based backlight includes a light guide configured to guide light and a plurality of the reflective micro-slit scattering elements having sloped reflective sidewalls configured to reflectively scatter out the guided light as the emitted light. The sloped reflective sidewalls of the reflective micro-slit scattering elements are configured to provide the predetermined light exclusion zone of the emitted light. The multiview display includes the reflective micro-slit scattering elements arranged as an array of micro-slit multibeam elements. The multiview display also includes an array of light valves to modulate the directional light beams to provide the multiview image, except within the predetermined light exclusion zone.

Abstract: A reflective microprism scattering element based backlight, a multiview display, and a method backlight operation include reflective microprism scattering elements configured to provide emitted light having a predetermined light exclusion zone. The reflective microprism scattering element based backlight includes a light guide configured to guide light and a plurality of the reflective microprism scattering elements having sloped reflective sidewalls configured to reflectively scatter out the guided light as the emitted light. The sloped reflective sidewalls of the reflective microprism scattering elements are configured to provide the predetermined light exclusion zone of the emitted light. The multiview display includes the reflective microprism scattering elements arranged as an array of reflective microprism multibeam elements.

Abstract: Feature importance is critical to understanding how predictive models produce accurate results, and can change significantly for different models. The present invention is used to achieve a good ranking for stable feature importance. An optimized technique is presented which considers feature importance value variation within different groups of cross-trained models. Feature importance is computed for all group models with this optimized method, and then a best set of models can be selected based on classification error as well as optimized stable feature importance values.

Abstract: A method for receiving an unreduced training data in the form of a plurality of TDRs (training data sets), using a first autoencoder on the plurality of TDRs to obtain first reduced training data in the form of a plurality of TDRs, clustering, into a plurality of clusters, the plurality of TDRs of the first reduced training data so that each TDR belongs to a cluster of the plurality of clusters, assigning a cluster identification code to each cluster of the plurality of clusters, adding cluster identification codes to each TDR of the plurality of TDRs of the unreduced training data, and perform stratified sampling to form K data blocks from the TDRs of the unreduced data.